Frequently Asked Questions

Does InsideForest replace my production model?

No. InsideForest starts from an already trained forest and transforms it into interpretable segments. Use it alongside production models to explain predictions.

Do I need labeled data?

Yes. InsideForest performs supervised clustering: it requires labels to guide the region search. Labels may come from experts, historical decisions, or other models.

Can it work with PySpark?

Yes. Use Trees('pyspark') to extract branches from a Spark RandomForestClassifier and continue with Regions and Labels just like in scikit-learn.

How do I interpret the rules?

Use explain_regions() or regions_ for region geometry and supervised metrics, and assign_regions() for row-level evidence. The older df_clusters_description_ table remains available to low-level legacy pipelines.

How large can my dataset be?

InsideForest has been tested with millions of rows on CPU. Enable auto_fast and follow Performance Tips to optimize resources.

How do I report issues?

Open a GitHub issue describing the environment (version, operating system, reproduction steps). Include pip freeze output and the result of pytest -q.

Where can I find full examples?

Explore the notebooks in the experiments/ directory and the published tutorials. You can also review the repository README.

Have another question? Join the discussion in the repository issues and contribute to the roadmap.